Services
What we do, and how it actually runs.
Six lines of work, three ways to engage. The first three are the specialisms — ATM and kiosk software, financial services, and the compliance and reconciliation reporting that goes into every system we ship.
ATM & kiosk software
Self-service software that has to work when nobody is standing next to it.
We build the application layer for physical ATMs and kiosks, integrated with Genmega hardware — dispensers, bill acceptors, receipt printers, scanners and encrypting PIN pads. Unattended devices are unforgiving: there is no operator to retry the step, and every failure that touches cash has to be explainable the following morning.
- Genmega dispensers, printers, scanners, acceptors and EPP
- A hardware bridge so the interface speaks one protocol, not six vendor SDKs
- Journalling before the dispense, reconciliation after it
- Versioned field deployment with a tested rollback path
- Diagnostics an on-site technician can actually read
Financial services & payments
Cash advance, POS debit and pre-authorisation, across attended and unattended channels.
Credit card cash advance, POS debit advance, pre-auth kiosks and pre-auth mobile apps, built on one platform so the record is identical wherever the transaction started. The transaction itself is the easy half — the work that decides whether the system is operable is the reconciliation and the reporting behind it.
- Credit card cash advance and POS debit advance
- Pre-authorisation flows on kiosk and on mobile
- One transaction record across every channel
- Daily settlement reconciliation with exception reporting
- Audit trail from origination through settlement
Compliance & reconciliation reporting
The reports that let you prove the system behaved — built into every system we ship.
This is not a module we sell separately; it is part of every build. Transaction-level audit trails, daily reconciliation against settlement, and exception reports that surface a difference instead of quietly absorbing it. If you cannot demonstrate what happened and that the numbers agree, the software is not finished.
- Transaction-level audit records, retained and queryable
- Daily reconciliation against processor and settlement files
- Exception reports that name the discrepancy and its origin
- Operator and regulator-facing report formats
- Designed in from the schema, not reverse-engineered before an audit
Applied AI engineering
AI that changes what the product can do — not a chat bubble in the corner.
We have shipped retrieval over private document sets, AI-assisted clinical workflow, and machine-facing APIs designed for agent callers from the start. That experience mostly teaches you where these systems fail: ungrounded answers, silent retrieval misses, and evaluation nobody set up until it was too late to argue with the output.
- Retrieval-augmented generation grounded in your own corpus
- Agent workflows that write to real systems, with an audit trail
- API surfaces designed for software callers, not scraped from a human UI
- Evaluation harnesses before rollout, so quality is measured not assumed
Custom product build
A system that did not exist, taken from idea to production.
Information architecture, interface design, data model, backend, authentication, deployment, certificates and the operational plumbing. One team accountable for whether it works, rather than a design studio, a dev shop and a DevOps contractor pointing at each other.
- Multi-tenant architecture designed in from the first schema
- Web, installable PWA, or native iOS and Android
- Containerised and reproducible so you can host it yourself
- Documentation written for the engineer who inherits it
Integration, infrastructure & rescue
The layer nobody demos, and the codebases nobody wants to inherit.
Event buses, idempotent APIs, outbox delivery, hash-chained audit and self-hosted deployment — the pattern we run in production ourselves. We also take on inherited systems: map what exists, stabilise what is on fire, then extend it, or tell you honestly when a rewrite costs less than another year of patching.
- Idempotent APIs that make duplicate events structurally impossible
- Outbox delivery so an offline consumer delays rather than loses data
- Docker on your own VPS or cloud, no vendor lock-in
- Codebase and infrastructure assessment with a written verdict
- Knowledge transfer, so you are not swapping one dependency for another
Engagement models
Pick the one that matches your uncertainty.
Discovery sprint
1–2 weeksA fixed-scope piece of work that ends in an architecture, a plan and a real number. Useful when the shape of the problem is still moving.
- Technical approach
- Scope and sequencing
- Costed plan you can take elsewhere
Build engagement
6 weeks – 6 monthsThe main event. Shipped in working increments you can see and use, rather than a six-month silence followed by a reveal.
- Working software each cycle
- Deployed environment from week one
- Handover documentation throughout
Ongoing partner
RollingWe keep building and keep operating: features, monitoring, upgrades, incident response. For teams who want a software department without hiring one.
- Continuous delivery
- Operational ownership
- Direct access, no ticket queue
Not sure which of these you need?
That is what the first conversation is for. Describe the situation and we will tell you which one fits — including when the answer is none of them.